Good afternoon, everyone, and thank you for joining me on our first day of our annual healthcare conference. My name is Gil Blum, and I am a senior biotech analyst here at Needham & Company. It is my pleasure to have with me today Zach Jonasson, the Chief Financial Officer of Absci Corporation. And we will be discussing some truly interesting developments in AI and its potential use cases in biology. With that, Zach, you have the floor.
Great. Thanks, Gil. I'm really pleased to be here and pleased to share some updates, a description of our company. I'm gonna kind of cycle through some slides here as backdrop. First, just some disclaimers. Then just to sort of set the stage, Absci is a data-first generative AI drug creation company, and we integrate AI with wet lab technologies in order to engineer better biologics faster. I'm gonna talk a little bit about how we do that and share a little bit about some of the programs that we're developing internally. But our platform, in short, allows us to design in functionalities that are important for meeting patient needs, both for best-in-class and first-in-class therapeutics.
Some recent success, successes, for the company, we were happy to sign two new, drug discovery partnerships, one with AstraZeneca and one with Almirall last year in Q4. We also hit our guidance for 10 new active programs, and that refers to programs that we're developing with partners. We've started talking about our internal asset portfolio in October of last year, and I'll be providing some updates today, particularly on the lead program, ABS-101, which is a TL1A antibody designed to be best in class. Just moving forward, a little bit about the platform to start off with. What's really unique about Absci is we have a wet lab technology that was built up over the early period of the company.
When the company first started, it was a synthetic biology company, and we leveraged this assay platform to generate large quantities of binding data, proprietary data that we use to train our AI models. We can generate datasets around millions of unique antibody sequences and how they interact with a given antigen's epitope in a single workflow. We combine that as well with public datasets to generate very large, massive-scale datasets that, again, we use to train our AI models on. Then we also deploy state-of-the-art AI architectural innovations that allow us to take that training data, develop models, and access vast sequence search space, so up to 20 to 55.
So we're really able to access and tap into this massive diversity of sequence space for antibodies as we zero in on the antibodies that provide the functionality that we're looking for. And then just a note about our wet lab. You can see in my background, I'm not actually in the lab today, but that's a picture of the lab behind me, as well as here on the slide. We have a 77,000 sq ft laboratory just outside of Portland, Oregon, where we run this entire cycle on a 6-week timeline. So generating data to train the models and then building the models, taking those designs that come out of the models and validating those in the wet lab. And I think the validation point here is really important.
We can validate up to 1 million unique antibody sequences in a given week. So not only are we building training data to build the AI models upon, we can actually validate the model predictions. And then we have a number of other validation tools that are lower throughput looking at hydrophobicity and some of the formulation attributes that we run in-house at Absci as well. So when we put the entire technology stack together, we also have the ability to start at target discovery. We have a novel target discovery platform, which we call reverse immunology, where we take patient samples, and we can identify novel antibodies produced by TLSs and then de-orphan the targets that they're after.
We can do that to find novel targets, or we can work with partners or select targets for fast follower that then go into our AI guided antibody drug creation platform, where we use de novo capabilities, to design best-in-class features into those antibodies. Those also then go through a final step of AI-guided lead optimization, where we use multiparametric optimization routines in order to make sure those antibodies are highly developable, removing any kind of bottlenecks or concerns that you would have around formulation, stability, and other parameters. We also use that process to design in unique functionalities, such as half-life extension, multivalency, and other properties that could be really differentiating for an antibody-based therapeutic. So what does all this mean? It's, it really, this platform has been designed to deliver unique value propositions.
Foremost is the ability to access unique and novel disease biology, and in fact, we're using our platform today to really go after a lot of novel, hard-to-address targets, such as GPCRs and ion channels, and we have had success there as well. So we're using the platform to access those targets. We're also using the platform to design in these unique functionalities. Secondly is increasing the probability of overall success. So this comes back to that multiparametric optimization AI. Here, we're making sure we engineer in properties that are gonna make the molecule be successful in both in terms of formulation, other developability, manufacturing features, non-immunogenic features, et cetera. Thirdly, using the platform, we can greatly reduce the time and cost to reach the clinic.
Right now, in our first lead program, ABS-101, it'll be an estimated two-year timeline to reach IND from start to finish, and cost Absci roughly $14 million-$16 million all in. So as you compare that to industry averages, at least for pharma, those timelines can be around five years, $50 million-$80 million. So in the same budget, Absci could develop 4-5 unique antibody therapeutics versus industry. So we are looking forward to leveraging that capability.
Then finally, I want to point out something that, I don't think gets talked about enough, but using our AI platform, and particularly the validation component in our wet lab, we're able to generate broad intellectual property, so a lot of composition of matter claims, and also able to navigate around other IP if in the case of a fast follower strategy. The primary business model that the company's pursuing is partnership-based, so we partner both at the drug creation phase, and here we've sort of highlighted our recent deals with AstraZeneca and Almirall. Those follow a typical drug creation or drug discovery partnership model, where there's R&D funding, upfront funding, and R&D funding, and then followed by clinical milestone funding points and commercial milestones, and then royalties after that.
As noted here in our two most recent deals, they tracked over $900 million in bio dollars, not including the royalty component. We also look to partner our internal assets at later stages of development, and we'll talk more about that in a minute. Just a note about kind of the ecosystem that we've created at Absci. Looking at the circle on the left-hand side, we have a number of partnerships that give us access to novel data sets, as well as compute. Our partnership with NVIDIA is highlighted here, and we've used those kinds of partnerships to help support what we've done internally for our own internal asset pipeline development. We currently have three internal programs, two of which are fast follower best-in-class, and a third, which will be a first-in-class program.
We'll talk a little bit more about those in subsequent slides. Then finally, we leverage these partnerships in this ecosystem as well to work with our partners, such as Merck, AstraZeneca, and Almirall. So now just a quick look at our internal pipeline. We have ABS-101 is our lead program. That is a program that's designed to be a targeted best-in-class for TL1A, and we'll talk a little bit more about that in the next few slides. Our second program, ABS-201, is a dermatology target. We've not disclosed that target yet, but we believe this is a highly underappreciated target class, and it's one where we would be second to market. And then a third program, which is a first-in-class that derives from our reverse immunology platform.
This is a immuno-oncology program, targeting the innate immune system, ABS-301, and we'll be releasing some more data on 201 and 301 later this year. But in the next slides, I'm gonna talk a little bit about ABS-101. So ABS-101 is targeting TL1A, which is a very well-validated target. It's involved in pro-inflammatory signaling, particularly in IBD, but also in fibrotic pathways. And the IBD application and indication's a very large market with a large amount of patient need, so we're looking forward to developing this asset into that market. We're also looking at the fibrotic markets, where we believe there are other large opportunities for this molecule.
And just a little bit about how we designed this program, and this is should give you some insight on how we've deployed the the AI platform I talked about earlier to design in features that we believe will make a best-in-class molecule for TL1A. So we've used our de novo AI here to design this molecule, particularly the CDRs for this antibody. And here we've designed in high affinity and potency, an extended half-life, which we believe will enable longer dosing intervals with a target of once quarterly. We've also used a multiparametric optimization to ensure that we have a very stable molecule suitable for sub-Q dosing and very good favorability or very favorable developability parameters as well, and then differentiated intellectual property. So I'll share a little bit of data here.
The first slide here on the left is looking at the affinity of three of our lead programs here. We have selected a lead that is now in IND-enabling studies, but this is data we presented here in January. Here you can see that we were able to create a very high-affinity mAb, much more, much higher affinity than the Merck asset, and on par with the Roivant. If you look at the chart on the right-hand side, this is a potency assay. We've done a number of these. Here we're just showing one of the assays. It's an apoptosis inhibition assay in TF1 cells.
Here you can see that our lead programs or our lead molecules have higher potency, significantly higher potency than the Merck molecule and slightly higher potency than the Roivant molecule. Then looking a little bit at what we've done on half-life, here on the left-hand panel is a slide showing increased recycling of FcRn, and this is an assay that we've run comparing our leads against the Roivant and the Merck molecules, and you can see much higher recycling. Then if you translate that over into looking at some PK studies we've done in Tg32 mouse model, here we're comparing one of our molecules to the Roivant, and you can see a much longer half-life.
We've subsequently also done a more robust set of PK studies in this mouse model, comparing to all of the other molecules ahead of us in the clinic with favorable outcome. Moving forward, one of the unique things about the platform we have that gives us a lot of discretion and enablement for new biology is our ability to select the epitope that we want to target. And so in this case, we were very purposeful in selecting an epitope region very similar to the Merck region, and we did that purposefully to avoid the immunogenicity issues that the Roivant molecule has shown in phase II. So in particular, we believe, and this is hypothesis, that the Roivant ADA response that was seen in phase II is likely driven by a complex formation between the antibody and the epitope that Roivant's targeting.
Whereas when you look at the Merck-Merck's data from phase II, it's a much cleaner ADA profile. So we purposefully targeted this area of the molecule, of, the TL1A receptor, in order to enable a much lower immunogenicity profile, particularly versus the Roivant molecule. So if we put all the data together, and we've just shown you a couple pieces here, what we've done is we've used the AI platform here to design what we believe will be a best-in-class TL1A antibody. We believe it'll have low immunogenicity based on the epitope we're targeting, high bioavailability. It'll be suitable for sub-Q auto-injector and a very high concentration formulation. And then we've engineered in a longer half-life, and as I mentioned, we're targeting a once quarterly dosing regime.
I think if we compare this across the programs that are ahead of us in the clinic, I think we're very well-positioned. Then just looking at some of the timelines here for this program, so in January we had completed all of our work, and then we initiated our IND-enabling studies for this program and selected a lead candidate in February of this year. We expect the IND-enabling studies to wrap up at the end of this year, potentially very early next year, and then we'll be in a position to file our IND in Q1 of 2025. We would then expect to quickly thereafter initiate a phase I trial, and our expectation is, in the latter half of 2025, we should have an interim data readout from that phase I trial.
Just taking a quick step back to look at the overall Absci organization. Currently, or at the end of the year, early January, we're at 160 employees full-time. I think what's really unique about Absci is, in terms of the organization, is our ability to recruit and attract expertise from different parts of the technology spectrum, including AI, synthetic biology, and drug discovery. We have talent coming from places like NVIDIA, Google, OpenAI, as well as from pharmaceutical companies like Shire, Takeda, and AstraZeneca. As I mentioned, we have a 77,000 sq ft facility where we do our wet lab-based, scalable wet lab-based assays that drive data creation and validation for our AI models, and then we've been successful at raising significant amounts of capital to support our mission.
Just a quick note about our team and board here. We, I know we're running short on time, so I'll just highlight a couple people on this slide. Number one, I'll highlight Andreas Busch, who's our Chief Innovation Officer. Andreas has 10 approved drugs under his belt. He was formerly the Head of Global R&D at Shire, and before that at Bayer, and he's been a key force in the company in sort of recruiting A-plus talent to run our discovery programs. And then I'd like to highlight, as one of the new board members, Sir Mene Pangalos joined Absci's board in January of this year. He comes from AstraZeneca, where he was the EVP of R&D, and brings a tremendous amount of expertise in the drug discovery and development arena.
In closing, I just want to say we're well on our mission to delivering better biologics faster and at lower cost, again, built on our wet lab capability as well as our AI capabilities. We're really excited about the internal pipeline that I discussed today. Just here are a couple catalysts that we're looking forward to over the next 12-18 months. Currently, the IND-enabling studies are underway, so we look to have those completed at the end of this year, early next year. We additionally look to have additional data releases around our other programs, ABS-201 and ABS-301, later this year. With that, I'll hand it back over to Gil.
All right. Thank you for that, Zach. So the call is now open for questions, but maybe I'll start on here. So, you know, there's quite a few names in the space right now, and they all claim to have a certain specific flavor. What do you think differentiates you from some of the other players in the space?
Yeah, Gil, it's a great question, and I would say... I'd go a step further and say there's a lot of noise in the space. I think there's a couple key things to point out here. Number one- just to level set, Absci's focused on large molecules, so antibodies. Most of the other players in the space are focused on small molecule. So that's a key difference right there, and we could talk a lot more about the difference in the type of AI and, and the challenges for small molecule versus antibody. But right there is a key difference. I think secondly, and, and probably the most important, is our data capability. So we're right now in the industry to develop drugs using AI, you're not really at a compute limitation.
The bottleneck is really around data, and we're not the only ones to talk about this. Daphne Koller gave an interview last week saying the same thing. So what Absci has brought to the table is an ability to generate data at scale, and it's functional data. So it's interrogating the antibody sequence or the antibody structure and its ability to interact with this specific epitope in doing that at scale. And that functional data really drives our ability to train AI models, as well as we use that same technology to then go and validate. And then the third thing I'm gonna point out is I think we have a world-class discovery team. So it's not just the great AI talent, the data, but really experienced discovery people that come out of Shire, Takeda, and other large pharma.
So we're leveraging that expertise to make sure we make the right selections on our internal programs.
Maybe focusing a little on the use case here for, for AI and machine learning. You really have the potential of optimizing literally any antibody towards any target. Have you guys given thought about, you know, what that broadly could imply for known antibodies and known targets, on the collaboration side? You know, think about, you know, extending IP and really getting better therapeutic windows from some of the agents out there.
Yeah, I mean, absolutely. I mean, as you'll note in our internal pipeline, two of the programs are past FIH. And TL1A is a great example where we see a target that's got validation. There's a big market in IBD, as well as secondary markets and other fibrotic indications. And the lead molecules in the programs have, I think, significant weaknesses. So we used our AI platform, in this case, to really design in to address those deficiencies, so we provided a much longer half-life. We've also introduced higher potency. We've designed it to go after an epitope where we'll have less immunogenicity. So those are the kinds of things that we're able to do and demonstrating with the platform today. But to your point, as we look forward, we're building in new functionalities into our AI models as well.
So the ability to look at multivalency, so targeting, having one antibody hit multiple targets, pH dependency, a number of other features that could really enable best-in-class.
Okay. Maybe from, you know, kind of the more business model for the company, do you see yourself more as a producer of novel molecules and targets for early clinical development, kind of with revenues being generated over time by partnerships, or more of internal pipeline development, or maybe, maybe a bit of a mix?
Yeah, I think mix is the right way to think about it. So, we have done and we always will look to do drug creation partnerships. And there, you know, part of the reason to do that is the platform is broad, so it allows us to diversify the types of indications we go into. Also, those partners bring a lot of expertise in the disease biology or the given target. So for example, we announced last year a deal with Almirall. Almirall is a pharmaceutical company that's purely focused in dermatology.
Mm.
We're able to leverage their expertise. They're leveraging ours as well. It's synergistic, but that gives us a way to diversify what we're doing and partner with companies that really are deep into the target or the disease biology. For our internal programs, we want to focus on areas that we can support. We can't, we can't do everything under the sun. So we're really looking at areas around cytokine biology, so that covers parts of immunology as well as oncology, where we're building up and have built up expertise. The idea there is we'll take those programs into human proof of concept, most likely phase I, but potentially a little bit later, and look to partner those for more significant deal economics.
Makes a lot of sense. But is there a situation in which you may consider, you know, really going all the way, to commercial stage, you know, like an ultra-rare indication or a rare indication, or really thinking about value creation on the phase I, phase II?
Yeah, you know, I never say never, but in this case, I would say very, very unlikely, right? And the reason is, you think about the efficiencies that our platform gives us. We can go from target to candidate extremely quickly. With TL1A, that was about a 14-month period. We believe that is continually gonna become less and less as the platform improves. So we can go really quick to get to candidate, and then we can go very quickly to get an IND. So our efficiency, you know, if I'm trying to maximize the return on investment for us, I think focusing on generating those assets, quickly getting to the point where we have enough validation to sign a very significant partnership with a company we believe who could drive that molecule into the commercial market, makes a lot of sense.
And then on the other side of the column, continuing to do the discovery partnerships as well.
Okay. And kind of a last one, and, you know, overall space question: Where do you think we're going in the next five years? There's a few of your peers out there, there's you out there.... Are we gonna see true application of AI in biotech? Do you, are you a believer?
Well, clearly, clearly I am. I wouldn't have joined Absci. I do think that it we're, you know, if you talk to people, in the AI space, just I meant broadly AI space, it's moving faster than anybody really expected. In the particular application area we're working in, really the limiter is data.
Yeah.
Right? And so our mission is to really ratchet up our data creation. And when I talk about data, some people sort of think, "Oh, just quantity of data." Really, data has to accommodate three dimensions. One is it has to be quality, so you have to be having data that's, like, qualitative and specific. You have to have quantity. But thirdly, it has to be usable.
Yeah.
You know, if you have petabytes of data on images, it's not clear all of that's usable. The kind of data we generate is the functional interaction data. We also marry it with public data, too, but that in-house data is really important. So where do I see us headed? I think that I see us getting better and better at developing, and faster and faster at developing, antibody-based therapeutics and maybe more exotic-type therapeutics to any given target, even those targets that are not addressable today. And I think over a longer time interval, you'll start to see AI make really significant inroads on the target discovery piece as well. That's a harder problem.
Yeah.
It takes a lot more data. There's a lot more nuance to that data. And I think that'll be a second wave that follows kinda what we're doing, where we can take a given target, where we have some confidence in the biology, whether it's novel or it's something with validation, and we can design a really unique antibody to deliver the therapeutic benefit that we're all looking for.
Yeah. I definitely think Absci has an excellent use case for this technology, and look forward to seeing what happens with the company next. Thank you, Zach, for your time.